A survey on Approaches and Problems in Automatic Ontology Generation

The automatic generation of ontology is a complete different process from the human based development. Indeed an automatic process must be able to undertake autonomous choices based on machine computable algorithms and processes to fill all those tasks that are normally hand made by knowledge engineers. Example of such tasks can be the selection of a well formed source corpus, large or not, from which extract conceptual knowledge, to deal with heterogeneous sources and formats, integrate them coherently, be able to organize extracted entities as part of the ontology consistently, generate well formed ontological representations, and even be able in somehow to validate the result. To cover this lack there are several possible approaches to be considered. Throughout this Chapter we firstly provide an overview of ontology phases required by an automatic system and organize them into an ontology generation life-cycle. Furthermore we analyze most of the current systems that specifically target the generation of an ontology, i.e. systems that given a non ontological source corpus are able to provide as output an ontology. Finally we provide an evaluation of these systems, their approaches, the resources involved and we conclude giving elements of comparison and highlight some of the existing issues that should be covered before to obtain a realistic implementation of a partial or complete automatic ontology generation system.